1. Kevin Hartnett has written a very interesting piece about the mistakes that AI makes when evaluating the content of a computer screen image.  This matters for many different scenarios such as automated vehicles.  Would a human make a different (and better) decision than the computer if the two are given the same information?  Hartnett's piece is based on this new academic study.  

    Hartnett writes:

    "Researchers are still trying to understand exactly why computer vision systems get tripped up so easily, but they have a good guess. It has to do with an ability humans have that AI lacks: the ability to understand when a scene is confusing and thus go back for a second glance."

    It appears that people are more humble than the computers.  The "arrogant" computers are too sure that they know the true data generating process.  This logic is identical to the discussion by Lars Hansen and Tom Sargent on robust decision making  .   Rational expectations theorists posited that we all agree on the correct data generating process. Robust decision making research retreats from this assumption as economic agents recognize that they "know that they do not know" the true data generating process. Such individuals seek out strategies to protect themselves against worst case scenarios even if they have trouble quantifying these risks.

    A humble decision maker recognizes when things "don't look right" and she seeks a 2nd opinion or at least another moment of reflection before making a costly decision.  

    My new book manuscript pursues this theme of self aware decision makers who build in some flexibility into their menu of choices because they recognize that they themselves are not sure about how the world around them is changing over time.

    Given that the computers are trained on backward looking algorithm,  they are more likely to make a mistake when confronted with an ambiguous setting.  This is a version of the "Lucas Critique", when the rules of the game change --- people reoptimize --- are the AI machines able to do so?  This may be the key test of the Alan Turing conjecture.






  2. This piece in The Chronicle is worth reading.  The physicist argues that academics have weak incentives to interact with the public.  Academics respond to these incentives by writing esoteric pieces that few can read and understand.   This piece raises the issue of how should academics interact with the public?   In New York City, my parents attend events where Paul Krugman speaks at CUNY.   These events sound impressive to me as hundreds of people attend to hear a discussion about a key issue such as President Trump's opposition to globalization.   I would like to attend such events and I often watch YouTube videos when Jim Heckman or Raj Chetty is giving a "big think" lecture.

    The challenge for academics in interacting with the public regards making "nuanced" statements.   There are very few cases in economics where a statement is "always true".  The context often matters.   In physics, if you drop an apple or a rock -- I believe they fall by the same speed. In economics, we are constantly learning about "heterogeneous treatment effects".  The same policy often has different consequences for different people, at different points in time in different places.  This means that there is plenty of scope to argue about whether  a given intervention will be effective.

     Consider raising the minimum wage to $15 an hour.   Most economists continue to believe that such a policy will have unintended consequences as firms substitute to capital rather than labor and give fewer opportunities to teenagers and those who have been incarcerated in the past.  But, as the recent research on monopsony makes clear  (and see Alan Krueger's thoughts here) --- there are specific cases when this policy can be defended. 

    This example highlights that economists need to state their assumptions before we can make a definitive claim.  These assumptions (often expressed using math in order to clearly use language that everyone understands) takes time to layout and such "footnotes" do not lend themselves to being presented in a public sphere.  The economist is viewed as "the expert" and is expected to pontificate.  While James Bond had a license to kill, an economist given "a license to pontificate" may be even more dangerous!

    My proposal here is to have a series of public debates.  A supporter of sharply raising the minimum wage should be on a panel with an opponent of the minimum wage.  Let each try to be witty and smart in front of the audience and have a moderator do her job to allow information to be generated and allow the adults in the room to them update their beliefs.

    In this age of Big Data, more and more economists use data to measure the size of different effects. For example, will unemployment rise by a little or a lot if the minimum wage goes up to $15 an hour?  More economists could play a key public role by being willing to discuss and debate evidence and statistical methods in public.  The challenge would be to strip the discussion of our arcane lingo (heteroskedasticity!  selection effects, instrumental variables, GMM, RD etc) without losing the intellectual content of the debate.

    The public's respect for economists would rise if we could take the fascinating fights that take place in the seminar room and bring them out into the open using a language that smart people who are not PHD economists could understand.  Such airing of "known unknowns" would build up our reputation as we compete with the other social sciences for influence and $.


  3. The New Yorker magazine has a very interesting new profile of Mark Z of Facebook.     This piece reminds me of my interest in the issue that Linkedin has been able to avoid all of the recent controversy that Facebook faces.  There are "firewalls" between these firms. Linkedin focuses on our "work life" while Facebook focuses on our "home and peer life".  Given that the modern company strives to be apolitical, perhaps it is no accident that there has been less "fake news" on that work platform? 

    To an economist, an interesting issue arises concerning network externalities.  The typical economist would argue that social media users would gain an increase in well being by integrating these two platforms into one mega-platform.  I have discussed this issue before in this blog post.    The economics of Internet "fire walls" would appear to be an under-researched topic.  By limiting the extent of "contagion" due to there being multiple-platforms, rumors may travel more slowly across the web. Having multiple platforms may thus offer a type of insurance against "contagion risk".

    In addition to having multiple platforms (and thus having firewalls), should social media posters pay a small fee for posting material?    In the 1990s, international finance economists would debate the merits of a Tobin Tax on transactions.  While such a tax would impose a deadweight loss , it would reduce the probability of bank runs and contagion.    In the case of posting to the Internet, should there be an equivalent Tobin Tax? Is the price of zero, too low of a price for sending an email (think of spam emails) or for posting social media content?  How would the volume and composition of social media posts be affected by such a tax?  Would the % of "fake news" rise or fall on Facebook?




  4. In West Los Angeles, young people zoom around on "the Bird". This electric scooter travels at about 10 to 15 miles per hour and is a substitute for walking, the bus or UberX.  As I understand it, Bird riders are supposed to ride while wearing a helmet but I estimate that less than 5% of riders actually use one.  So, this is the opposite case of the "Peltzman effect".  In his 1975 JPE paper, Peltzman discusses the consequences of mandatory seat belts for driving behavior.   The same issue arises for mandatory airbags for cars.  When people feel safer behind the wheel, they change their behavior and drive faster and this has the unintended consequence of injuring more pedestrians.

    In the case of the "Bird", everything is reversed.  A rider who does not have a helmet is aware of the laws of physics and has strong incentives to protect her head.  So, a good economics paper could be written on whether there have been fewer nasty Bird incidents than the engineers would have predicted.  This would be evidence of behavioral modification (i.e safety precautions). 


  5. Since August 2017, I have been working hard as the Chairman of USC Economics.  While certain anonymous websites like to beat up on USC,  the department is hiring a new cohort of really good young faculty and we are now ranked #18 on REPEC.   Perhaps to celebrate this progress, I was invited to attend yesterday's first football game and to meet some of the University's trustees and the Provost.

    In the middle of the game, I was thrilled to see this.


    Image may contain: Matthew E. Kahn, crowd, stadium, sky and outdoor
My Research and My Books
My Research and My Books
To learn more about my research click here.

To purchase one of my four books, click here.
Popular Posts
Popular Posts
Blog Archive
Blog Archive
About Me
About Me
Loading
Dynamic Views theme. Powered by Blogger. Report Abuse.